Enable remote agent on Windows AArch64
Categories
(Remote Protocol :: Agent, defect, P1)
Tracking
(firefox74 fixed)
Tracking | Status | |
---|---|---|
firefox74 | --- | fixed |
People
(Reporter: ato, Assigned: ato)
References
(Blocks 1 open bug)
Details
(Whiteboard: [puppeteer-beta-mvp])
Attachments
(4 files)
Bug 1543115 disabled the remote agent on Windows AArch64
due to a dependency on winapi 0.2.8 which doesn’t support AArch64.
The dependency chain is http
0.1.17 → bytes
0.4.9 → iovec
0.1.2 → winapi
0.2.8.
We should upgrade/submit upstream PRs to move away from winapi
0.2.8.
(See attachment for the full dependency tree.)
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Updated•5 years ago
|
Comment 1•5 years ago
|
||
Please note that winapi is already on 0.3.8, and the necessary patches for ARM landed in 0.3.7, which is already vendored as third_party package.
iovec: v0.1.3 released on Oct 9th which got rid of the winapi dependency (https://github.com/carllerche/iovec/pull/25)
bytes: iovec has been replaced with std::io::IoSlice on June 7th, which hasn't been released yet. That is mostly because it waits on the tokio 0.2 release, which doesn't seem to happen soon.
Assignee | ||
Comment 2•5 years ago
|
||
If we upgrade to http
0.2, this would give us bytes
0.5 which contains the removal of the iovec that depended on the incompatible winapi
.
If my thinking is correct this would allow us to build on Windows AArch64 and simulatenously solve bug 1606935.
Assignee | ||
Updated•5 years ago
|
Updated•5 years ago
|
Assignee | ||
Comment 3•5 years ago
|
||
Upgrading to http 0.2 will give us bytes ^0.5 which removes
a dependency on iovec. iovec is problematic for us because it
depends on a version of winapi that does not compile on Windows AArch64.
The effect of this should be that the remote agent is now possible
to compile on Windows AArch64.
Assignee | ||
Comment 4•5 years ago
|
||
Assignee | ||
Comment 5•5 years ago
|
||
The remote agent used to not compile on Windows AArch64 due to iovec's
dependency on a version of winapi without support for this architecture.
Now that the remote agent has upgraded to http 0.2, which depends on
a version of the bytes crate that has moved away from iovec in favour
of std::io::IoSlice, we are able to turn on support for Windows AArch64.
This in turn will also fix bug 1606935 because the browser-chrome
test manifest for M(remote) will no longer be empty. It was a
regression caused by 1603930 where we fixed a logic error causnig
ENABLE_REMOTE_AGENT to be inappropriately set on non-trunk branches.
Comment 7•5 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/bca976fbc2c1
https://hg.mozilla.org/mozilla-central/rev/2965d6589d90
https://hg.mozilla.org/mozilla-central/rev/4558b24acb30
Description
•